ChatGPT and Claude.
KAZUKIN is currently researching these two generative AIs at the same time.
Both are very capable.
Throw them a question and they read my intent, write text, come up with ideas, and sometimes present perspectives I'd never have thought of.
But even when asking the same question, the answers often differ quite a bit.
Of course, this is just KAZUKIN's personal impression from using them.
GPT cheers me on, Claude makes me stop and think
ChatGPT has a strong tendency to stay close to my opinion.
When I consult with “This is what I think—how about it?” it often gives me a push with things like “That way of thinking is sound,” “I think that judgment is fine,” and “It fits KAZUKIN's direction.”
I get the feeling: you're not wrong. Keep going.
It's nice. Hardly anyone feels bad being affirmed.
Even though I'd been unsure, by the time I finish reading the answer, there's a KAZUKIN who stands a little taller: “Right. I'm not wrong after all.”
Claude is a little different. It doesn't flatly deny my opinion.
But with things like “That way of thinking isn't bad. However, some people think this way,” and “From another perspective, this kind of problem could occur,” it calmly sets the fact that different opinions exist on the table.
The impression is of quietly gripping the shoulder of a giddy KAZUKIN and saying, “Well, let's sit down and think it over.”
ChatGPT, the cheerleader.
Claude, calm and composed.
That is how they look to KAZUKIN.
Having the AIs hold a debate
And when the two answers differ greatly, KAZUKIN uses a method.
I show ChatGPT's answer to Claude and ask, “ChatGPT says this—what's your opinion, Claude?” Then I take Claude's answer to ChatGPT and ask, “Claude says this—what do you think?”
A debate between AIs, with the person absent, begins. It's surprisingly fun.
I wondered if they'd stubbornly keep insisting they're right, but not really.
Sometimes they readily admit fault: “That point is valid,” “I lacked understanding of the premises,” “On this point, Claude's view is more appropriate.”
They're right. My answer had a gap. Given that premise, the judgment changes.
In a situation that might get awkward between humans, the AI corrects itself calmly. Pretty clean.
Multiple perspectives gather
Going back and forth between ChatGPT's and Claude's opinions like this, multiple views gather around a single question.
- An opinion that gives me a push
- An opinion that makes me stop and think
- Risks I'd overlooked
- Problems seen from another standpoint
If I'd used only one, there are probably things I wouldn't have noticed.
両方に聞いちゃうんだよね・・
Summary
But in the end, it's not the AI that decides.
Not because ChatGPT cheered me on. Not because Claude calmly cautioned me.
After hearing both opinions, understanding the differences, and lining up benefits and risks—so what will KAZUKIN do?
In the end, we come back here.
AI is an excellent advisor. But it won't press the life-decision button for you.
Gather various answers, think, and choose for yourself in the end. That is probably what matters.
Even so, when in doubt, I ask both again today. And if the answers differ, I toss their opinions at each other.
Watching their discussion, KAZUKIN thinks once more.
They give pretty different answers, huh...
KAZUKIN Rating
- 😊 How ChatGPT gives a push — ★★★★★
- 🧊 Claude's calm pointing-out — ★★★★★
- ⚔️ Fun of AI-versus-AI debates — ★★★★★
- 👀 Rate of discovering overlooked risks — ★★★★☆
- 🙋 Ability to decide for myself in the end — ★★★☆☆ (needs training)
This article is a personal opinion. KAZUKIN LAB has no relationship with OpenAI or Anthropic. The tendencies of the AIs' answers are just KAZUKIN's personal impressions from using them. All trademarks belong to their respective owners.